Fetch and Analyze YouTube Video Transcripts

The YouTube Transcript skill fetches transcripts, search results, and playlist or channel data from YouTube through the getyoutubetranscript.com REST API, so an agent can summarize, quote, search, or analyze a video's actual spoken content without the user copy-pasting it in by hand. It requires no yt-dlp and no Google API key. The skill is maintained as a community skill under the source repo tubeagentkit/youtube-transcript-skills.

Every call needs an API key, either an existing YOUTUBE_TRANSCRIPT_API_KEY environment variable or a key the user has already provided. If neither exists, the skill can walk the user through a free-tier signup (100 credits, no card required) via a two-step email and verification-code flow - but only after the user has explicitly consented and supplied an email address to use for that purpose.

The base URL for every call is https://getyoutubetranscript.com/api/v1, and the key is sent as an Authorization: Bearer header or as an x-api-key header. Errors come back as JSON with a stable code (for example TRANSCRIPT_NOT_FOUND, PAYMENT_REQUIRED, or RATE_LIMITED) and a matching HTTP status; the skill is meant to relay the message field to the user rather than guess at a fix. Paginated endpoints return an opaque continuation_token, which is passed back verbatim as continuation to get the next page - null means there are no more pages.

curl -s "https://getyoutubetranscript.com/api/v1/transcript?v=jNQXAC9IVRw&language=en" \
  -H "Authorization: Bearer $YOUTUBE_TRANSCRIPT_API_KEY"

Use this skill when the user wants a YouTube video's transcript fetched, summarized, quoted, or analyzed; when they want to search YouTube globally or within one channel by handle; when they want a channel handle resolved to a channel ID, or a channel's info and upload history; or when they want the videos in a YouTube playlist.

Do not expect per-line timestamps - the transcript endpoint returns the full spoken text as a single string, with no per-line timestamp breakdown, so this skill is the wrong fit for a job that needs timestamp-aligned captions. It also does not replace environment-specific validation, testing, or expert review, and it should stop and ask for clarification rather than proceed if required inputs, permissions, or safety boundaries are missing. On a 402 PAYMENT_REQUIRED response, the right move is to direct the user to top up rather than retry, and a 429 RATE_LIMITED response should be backed off from, not hammered in a retry loop.

Inputs vary by endpoint: a video ID or URL and an optional language for a transcript (GET /transcript?v=&language=en); a query and type of video or channel for search (GET /search?q=<query>&type=video|channel); a channel handle for resolving a channel ID (GET /resolve?handle=@name, free), fetching channel info and latest videos (GET /channel/latest?channel=@name, free), or listing a channel's full upload history (GET /channel/videos?channel=@name, paginated); a channel handle plus a query for searching within a channel (GET /channel/search?channel=@name&q=<query>, paginated); and a playlist ID or URL for playlist videos (GET /playlist?list=<playlist ID or URL>).

Outputs are the corresponding JSON payloads: transcript text as one string, search results, resolved channel IDs, channel info and video lists, or playlist video lists, plus an opaque continuation_token on paginated responses. On failure, the output is a JSON error object with a stable code and message rather than a raw exception.

The skill's only outbound integration is HTTPS requests to getyoutubetranscript.com endpoints, plus, during first-time setup, an email address the user explicitly provides for the signup flow. It runs no other shell commands and installs nothing. A newly issued API key should never be persisted to a shell profile or other file without the user's explicit confirmation first. A related but separate skill, @youtube-full, offers an alternative YouTube transcript, search, channel, and playlist capability backed by a different upstream API.

This skill is for anyone building or running an agent that needs to read YouTube content programmatically - summarizing a video's spoken content, searching YouTube or a specific channel, resolving a channel handle, listing a channel's uploads, or pulling a playlist's videos - without asking the user to copy-paste a transcript by hand, and without needing yt-dlp or a Google API key.